summaryrefslogtreecommitdiff
path: root/src/bin/pg_dump/README.scrappy
diff options
context:
space:
mode:
Diffstat (limited to 'src/bin/pg_dump/README.scrappy')
-rw-r--r--src/bin/pg_dump/README.scrappy23
1 files changed, 23 insertions, 0 deletions
diff --git a/src/bin/pg_dump/README.scrappy b/src/bin/pg_dump/README.scrappy
new file mode 100644
index 0000000000..9a71008f75
--- /dev/null
+++ b/src/bin/pg_dump/README.scrappy
@@ -0,0 +1,23 @@
+ Here is what Marc had to say about insert.patch included in this archive....
+
+ In preparation of finally moving all my 1.0 databases over to a 1.01
+ database server, I looked at pg_dump and found that, unless I missed
+ something, it didn't *easily* do what I wanted, which was to dump a database
+ to a file, and then reload it again on another server (short-term)...but,
+ also, there doesn't seem to be any mechanism for dumping the database to a
+ file that can be backed up and quickly reloaded again.
+
+ So, I spent the past several hours modifying pg_dump so that it has an extra
+ switch for dumping the data in valid 'insert' strings, so that you can
+ quickly and easily reload a database.
+
+ So, now the output looks like:
+
+ CREATE TABLE scrap (integer int4, real float4, text text) archive = none;
+ insert into scrap values (1, 1, 'text');
+
+ Now, the hard part was figuring out what types are available, so that the
+ insert string works properly for char vs numberic data fields. As such, the
+ switch statement I'm using in dumpClasses() for this may be missing values
+ for numeric fields (I'm using PQftype() to figure out numeric vs non-numeric
+ fields)